Abstract
Social scientists have overwhelmingly documented a strong and increasing educa-tional homogamy between spouses. When estimating sorting by education, the pres-ence of measurement error in the education variables or random factors in the matching process may underestimate the actual degree of assortative mating, simultaneity bias may overestimate it, while omitting other individual characteristics relevant in the marriage market may under- or overestimate it. We address these issues using an in-strumental variables approach based on exploiting genetic variation in polygenic scores and controlling for population stratification. Specifically, we instrument spousal edu-cation with his/her educational polygenic score while controlling for own educational polygenic score. If the exclusion restriction is satisfied, our findings indicate that (1) assortative mating is underestimated when using OLS, and that (2) male education is correlated with other matching-relevant socioeconomic characteristics, while female education is productive per se in the matching. If the exclusion restriction is not satisfied, our evidence is consistent with (2). This suggests that individual socioeco-nomic attractiveness in the marriage market is multidimensional for men, but can be summarized with education for women.